Dear Alex,
> is there a way to compute main effects of study in SPM96? I tested two
> groups of subjects according to the following design (A, B, C are
> different tasks)
>
> Group1: scans
> 12 subjects A A B B
> 5 subjects A A C C
>
> Group2:
> 12 subjects A A B B
> 5 subjects A A C C
>
> naturally, i'm interested in overall as well as task-specific
> differences between groups. but entering this design as a multi-study
> with replications (e.g. as 4 studies) does not seem to help getting at
> the main effects of study since it would require that contrast weights
> between the two groups be of opposite signs (e.g. negative weights for
> all conditions of group 1 and positive weights for all conditions of
> group 2). This, however, is not a valid contrast (at least not in
> SPM96), since valid contrasts must sum to zero within a single study.
>
> Is there a way out of this? How would you suggest to best enter this
> study into SPM96?
You would simply combine all your scans from each subject into a single
scan and compare groups (this would represent the appropriate random
effects analysis).
I hope this helps - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|